John Hayward writes:
> 
>    After I label the disk and in the process of doing a newfs it proceeds
> to print the backup super blocks then stops with the following error
> message:
> write error: 0
> wtfs: Read-only file system
> 
> Follwing that the disk label is messed up.
> 
> Mike Dowden suggested a low level format and I found that test 75 on the 
> 3100 console provided for formatting a scsi drive.  This was performed
> successfully - so I believe the hardware is ok.
> 
> Is anyone else seeing these kind of problems?

Yes!  With my Vaxstation 3100/M38 and an external IBM 1GB drive.
With a kernel compiled on June 29 it works, with a new kernel it reports
no label and data read from the drive is "goofy" and the drive can't be
labeled.

I think some is "Bad Wrong(TM)" somewhere.  I guess I need to do a
binary search on it.  I pretty much know the kernel built on June 29 works
and a kernel around August 3 does not.  Time to search. :(

> I note that 1632*15*53 = 1297440 where the dmesg indicates 1299174
> I don't know if that is because of bad sector forwarding or different
> number of sectors per track over the disk or something else.

I see something like this also.  I have 4119*5*99 = 2038905 while the
dmesg reports 2055035.  I used 2038905 to create a 'sd3e' partition.
This has been working great ever since the SCSI was fixed and DMA was
added, etc.  In the July timeframe something seems to have broken.
I'll start looking at the CVS logs and try to checkout a specific date
and build a kernel.

-Andrew